Recently, next-generation mobile communication systems have been actively studied. As a method for enhancing the frequency utilization efficiency of a system, a single-frequency reuse cellular system has been proposed in which each cell uses the same frequency band so that each cell can use the entire band allocated to the system.
OFDMA (Orthogonal Frequency Division Multiple Access) is a front-runner in a downlink (communication from a base station device to a mobile station device). OFDMA is a communication system in which information data are modulated by use of different modulation schemes, such as 64 QAM (64-ary Quadrature Amplitude Modulation) and BPSK (Binary Phase Shift Keying), according to reception conditions to generate an OFDM signal, and radio resources defined by time-and-frequency axes are flexibly allocated to multiple mobile station devices.
Since an OFDM signal is used in this case, a PAPR (Peak to Average Power Ratio) becomes greatly high in some cases. The high peak power is not a severe problem for downlink communication having a relatively-high transmission-power amplifying function. However, the high peak power is a fatal problem for uplink communication (from the mobile station device to the base station device) having a relatively-low transmission-power amplifying function.
For this reason, a single-carrier-based communication scheme with a low PAPR is suitable to the uplink (communication from the mobile station device to the base station device).
However, the use of the single-carrier scheme causes a problem that flexible resource allocation using time-and-frequency axes cannot be performed such as in the case of OFDM. To solve the problem, SC-ASA (Single Carrier-Adaptive Spectrum Allocation), which is also called DFT-S-OFDM (Discrete Fourier Transform-Spread OFDM), has been proposed (see, for example, Non-Patent Document 1).
Such a communication scheme uses the same scheme as the single-carrier communication scheme, resulting in a lower PAPR. Additionally, a cyclic prefix is inserted as in the case of OFDM signals, enabling data processing without inter-block interference (in this description, an interval at which a cyclic prefix is inserted, i.e., data processing unit by which DFT is performed, is called a DFT-S-OFDM symbol). Further, frequency waves are generated once by use of DFT, thereby simplifying resource control per subcarrier.
FIG. 40 illustrates a configuration of a transmission device when MIMO (Multi-Input Multi-Output) transmission using SC-ASA is performed. FIG. 40 may be regarded as illustrating one transmission device including multiple transmission systems or as illustrating different transmission devices. This respect is explained hereinafter. In FIG. 41A, one base station wirelessly communicates with two mobile stations. Each of the base station and the mobile stations includes two antennas. If the configuration of the transmission device shown in FIG. 40 is regarded as one transmission device including multiple transmission systems, FIG. 40 is regarded as illustrating a case of single-user MIMO shown in FIG. 41C. If the configuration of the transmission device is regarded as different transmission devices, FIG. 40 is regarded as illustrating a case of multi-user MIMO shown in FIG. 41B. Subcarriers to be used are denoted as white blocks. Subcarriers corresponding to the numbers, which are not denoted as white blocks, are ones not selected in SC-ASA.
Regarding each transmission system shown in FIG. 40, transmission data 1 and 2 are encoded by encoders 1000 and 1001, and then modulated by modulators 1002 and 1003, respectively. The modulated signals are converted into parallel signals by S/P (Serial/Parallel) converters 1004 and 1005, and then converted into frequency-domain signals by DFT units 1006 and 1007, respectively. Spectral mapping units 1008 and 1009 perform mapping such that the transmission data 1 and 2 use the same frequency subcarriers as shown in FIGS. 41B and 41C. Subcarriers having high received SNR or SINR are used in the case of SC-ASA. However, MIMO transmission causes signals transmitted from the two transmission systems to interfere with each other on the receiving side. For this reason, common subcarriers having good conditions have to be selected for the transmission antennas (users) in consideration of the degree of the interference and all channels among the two transmission systems and the two reception systems.
Then, the mapped frequency-domain signals are converted into time-domain signals by IDFT units 1010 and 1011, and then converted into serial signals by P/S (parallel/signal) converters 1012 and 1013. Then, cyclic prefixes are inserted by CP inserters 1014 and 1015, and then converted into analog signals by D/A converters 1016 and 1017. Finally, the analog signals are upconverted into radio-frequency signals by radio units 1018 and 1019, and then transmitted from the transmission antennas 1020 and 1021.
FIG. 42 is a schematic block diagram illustrating a configuration of a reception device receiving signals transmitted by the MIMO systems. Since the reception device shown in FIG. 42 includes a canceller, the reception device having such a configuration can achieve better reception characteristics. The device shown in FIG. 42 includes: antennas 1100 and 1101; RF units 1102 and 1103; A/D converters 1104 and 1105; CP removers 1106 and 1107; S/P converters 1108, 1109, 1133, and 1134; DFT units 1110, 1111, 1116, 1117, 1135, and 1136; channel estimators 1112 and 1113; a canceller 1114; a signal equalizing-and-demultiplexing unit 1115; a spectral demapping unit 1118; IDFT units 1119, 1120, 1138, and 1139; P/S converters 1121 and 1122; demodulators 1123 and 1124; decoders 1125 and 1126; repetition controllers 1127 and 1128; determining units 1129 and 1130; replica generators 1131 and 1132; a spectral mapping unit 1137; and a channel multiplier 1140.
The signals transmitted from the transmission device shown in FIG. 40 are received by the antennas 1100 and 1101 of the reception device, downconverted from radio-frequency signals by the RF units 1102 and 1103, and then converted into digital signals by the A/D converters 1104 and 1105. Then, the cyclic prefixes CP (GI) added on the transmitting side are removed by the CP removers 1106 and 1107. Then, the signals with the cyclic prefixes removed are converted into parallel signals by the S/P converters 1108 and 1109, and then converted into frequency-domain signals by being subjected to DFT performed by the DFT units 1110 and 1111. Channel estimation between each transmission antenna and each reception antenna is performed using a known signal added on the transmitting side as a signal for channel estimation, the known signal being included in the converted frequency-domain signal. In this case, channel estimation values for the number of subcarriers are calculated with respect to 4 channels=the number of transmission antennas×the number of reception antennas.
The data signals subjected to DFT and then converted into the frequency-domain signals are input to the canceller 1114. The canceller 1114 subtracts, from the received signals, replicas of received signals, which are generated based on the reliability of demodulated data. If a perfect replica (transmitted signal) is generated, an output of the canceller 1114 includes only noise elements. This calculation can be expressed as an expression (100) where R denotes a reception-data vector received by the two antennas, Ξ denotes a channel matrix, and S′ denotes a replica of a transmission-data vector (generated by a replica generator to a spectral mapping unit as will be explained later).Q=R−ΞS′  (100)
Q denotes a vector indicating an output of the canceller 1114 at the time of second-or-more repeated operation (i.e., a residual after cancelling). R, Ξ, S′ are shown in the following expressions (101) to (103), where a figure in a parenthesis denotes the subcarrier number, and an index denotes the transmission-and-reception antenna number. Two indexes of Ξ denote a combination of reception-and-transmission antennas. For example, Ξ21 denotes a channel from the transmission antenna 1 to the reception antenna 2. These expressions may be used for both single-user MIMO and multi-user MIMO.
                    R        =                  [                                                                                          R                    1                                    ⁡                                      (                    1                    )                                                                                                                                            R                    1                                    ⁡                                      (                    2                    )                                                                                                                                            R                    1                                    ⁡                                      (                    3                    )                                                                                                                                            R                    1                                    ⁡                                      (                    4                    )                                                                                                                                            R                    2                                    ⁡                                      (                    1                    )                                                                                                                                            R                    2                                    ⁡                                      (                    2                    )                                                                                                                                            R                    2                                    ⁡                                      (                    3                    )                                                                                                                                            R                    2                                    ⁡                                      (                    4                    )                                                                                ]                                    (        101        )                                Ξ        =                  [                                                                                          Ξ                    11                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                                  Ξ                    12                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                    0                                                                                  Ξ                    11                                    ⁡                                      (                    2                    )                                                                              0                                            0                                            0                                                                                  Ξ                    12                                    ⁡                                      (                    2                    )                                                                              0                                            0                                                                    0                                            0                                            0                                            0                                            0                                            0                                            0                                            0                                                                    0                                            0                                            0                                                                                  Ξ                    11                                    ⁡                                      (                    4                    )                                                                              0                                            0                                            0                                                                                  Ξ                    12                                    ⁡                                      (                    4                    )                                                                                                                                            Ξ                    21                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                                  Ξ                    22                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                    0                                                                                  Ξ                    21                                    ⁡                                      (                    2                    )                                                                              0                                            0                                            0                                                                                  Ξ                    22                                    ⁡                                      (                    1                    )                                                                              0                                            0                                                                    0                                            0                                            0                                            0                                            0                                            0                                            0                                            0                                                                    0                                            0                                            0                                                                                  Ξ                    21                                    ⁡                                      (                    4                    )                                                                              0                                            0                                            0                                                                                  Ξ                    22                                    ⁡                                      (                    4                    )                                                                                ]                                    (        102        )                                          S          ′                =                  [                                                                                          S                    1                    ′                                    ⁡                                      (                    1                    )                                                                                                                                            S                    1                    ′                                    ⁡                                      (                    2                    )                                                                                                      0                                                                                                          S                    1                    ′                                    ⁡                                      (                    4                    )                                                                                                                                            S                    2                    ′                                    ⁡                                      (                    1                    )                                                                                                                                            S                    2                    ′                                    ⁡                                      (                    2                    )                                                                                                      0                                                                                                          S                    2                    ′                                    ⁡                                      (                    4                    )                                                                                ]                                    (        103        )            
The reason that replicas (ΞS′) of all signals including desired signals to be extracted are cancelled is that the signal equalizing-and-demultiplexing unit 1115 that will be explained later performs an inverse matrix calculation, and therefore the inverse matrix calculation has to be performed a number of times corresponding to the number of desired signals included in a block if cancelling and equalization are repeated without cancelling the desired signals. On the other hand, if the residual Q after the canceling of all replicas is input, the residual can be equally treated in the block, and therefore all weights can be calculated with one inverse matrix calculation with respect to the block. For this reason, the replica is independently input and reconfigured to decrease the amount of the inverse calculation. However, a replica of the firstly received signal cannot be generated. In this case, the reception-data vector (R) passes through the canceller 1115 as it is.
The signal output from the canceller 1114 is input to the signal equalizing-and-demultiplexing unit 1115, and then subjected to equalization using frequency-domain signals. When the repeated operation is performed, the signal equalizing-and-demultiplexing unit 1115 performs, with use of an expression (104), MMSE equalization on each signal generated by a replica of the received signal for each data vector transmitted from the antennas 1 and 2 shown in FIG. 43 being added to the output (Q) of the canceller. FIG. 43 illustrates, as an example of subcarrier selection, a case where subcarriers 1, 2, and 4 are transmitted from the antennas 1 and 2.z=(1+γTnδTn)−1[γTns′Tn+FHΨTnQ]  (104)
Tn (n=1, 2 in the above case) denotes a transmission antenna. γTn and δTn denote real numbers used when tap coefficients are calculated. Similarly, ΨTn denotes a complex square matrix having a size of the DFT block length, which is used when tap coefficients are calculated. s′Tn denotes a replica of the signal transmitted from the antenna Tn. Q denotes a result (residual) of subtracting replicas of all the received signals from the received signals. Since a replica of a received signal cannot be generated (s′Tn is a zero vector) in the first operation, the signal R output from the canceller 1114 without being subjected to subtraction is subjected to equalization. When calculating ΨTn and the like shown in the expression (104), channel matrices ΞT1 and ΞT2 corresponding to the transmission-data vectors 1 and 2 are used in addition to the channel matrix shown in the expression (102). ΞT1 and ΞT2 are channel matrices for respective transmission antennas, which are used for equalizing the transmission-data vectors 1 and 2.
                              Ξ                      T            ⁢                                                  ⁢            1                          =                  [                                                                                          Ξ                    11                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                    0                                                                                  Ξ                    11                                    ⁡                                      (                    2                    )                                                                              0                                            0                                                                    0                                            0                                            0                                            0                                                                    0                                            0                                            0                                                                                  Ξ                    11                                    ⁡                                      (                    4                    )                                                                                                                                            Ξ                    21                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                    0                                                                                  Ξ                    21                                    ⁡                                      (                    2                    )                                                                              0                                            0                                                                    0                                            0                                            0                                            0                                                                    0                                            0                                            0                                                                                  Ξ                    21                                    ⁡                                      (                    4                    )                                                                                ]                                    (        105        )                                          Ξ                      T            ⁢                                                  ⁢            2                          =                  [                                                                                          Ξ                    12                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                    0                                                                                  Ξ                    12                                    ⁡                                      (                    2                    )                                                                              0                                            0                                                                    0                                            0                                            0                                            0                                                                    0                                            0                                            0                                                                                  Ξ                    12                                    ⁡                                      (                    4                    )                                                                                                                                            Ξ                    22                                    ⁡                                      (                    1                    )                                                                              0                                            0                                            0                                                                    0                                                                                  Ξ                    22                                    ⁡                                      (                    1                    )                                                                              0                                            0                                                                    0                                            0                                            0                                            0                                                                    0                                            0                                            0                                                                                  Ξ                    22                                    ⁡                                      (                    4                    )                                                                                ]                                    (        106        )            
By the equalization with use of the expression (104), the equalized time-domain signals are output from the signal equalizing-and-demultiplexing unit 1115 for each transmission data (see, for example, Non-Patent Document 3).
The signals that have been transmitted from the respective transmission antennas and equalized for the respective signals are input to the DFT units 1116 and 1117, converted into frequency-domain signals by the DFT units 1116 and 1117, and then input to the spectral demapping unit 1118. The spectral demapping unit 1118 performs demapping common to spectra transmitted from the antennas 1 and 2 based on spectral mapping information. Then, the demapped signals are converted into time-domain signals by the IDFT units 1119 and 1120, converted into serial signals by the P/S converters 1121 and 1122, and then subjected to demodulation and decoding.
The demodulators 1123 and 1124 calculate LLRs (Log Likelihood Ratios) indicative of the reliability of the reception data subjected to error coding. The decoders 1125 and 1126 perform error correction decoding on the LLRs to update the LLRs. The repetition controllers 1127 and 1128 receiving the LLRs determine whether or not the repeated operation has been performed the predetermined number of times. If the repeated operation has been performed the predetermined number of times, the repetition controllers 1127 and 1128 output the LLRs to the determining units 1129 and 1130. On the other hand, if the repeated operation has not yet been performed the predetermined number of times, the repetition controllers 1127 and 1128 output the LLRs to the replica generators 1131 and 1132, and proceeds to a process of generating replicas of received signals. Assuming that a CRC (Cyclic Redundancy Check) is used, the repeated operation may end if no error is detected.
The replica generators 1131 and 1132 generate signal replicas (replicas of transmitted signals) corresponding to the respective LLRs. The generated replicas are passed through the S/P converters 1133 and 1134, and then converted by the DFT units 1135 and 1136 into frequency-domain replicas of signals transmitted from the respective antennas.
Then, the frequency-domain signal replicas generated in this manner are mapped by the spectral mapping unit 1137 based on mapping information received from a spectrum determining unit (not shown) in a similar manner as done on the transmitting side. Then, the replicas S′ subjected to the spectral mapping are input to the channel multiplier 1140, and then input to the signal equalizing-and-demultiplexing unit 1115 through the IDFT units 1138 and 1139. The signal equalizing-and-demultiplexing unit 1115 receiving the replicas S′ subjected to the spectral mapping reconfigures the received signals of the transmission-data vectors 1 and 2 using the replicas as explained above, and uses the reconfigured received signals for equalizing the respective transmission-data vectors. To generate replicas of the received signals to be used for subtraction from the received signals performed by the canceller 1114, the channel multiplier 1140 multiplies the replicas subjected to the spectral mapping by the channel matrix (Ξ shown in the expression (102)). Then, the replicas (ΞS′) of the received signals, which are output from the channel multiplier 1140, are input to the canceller 1114, and then subtraction shown in the expression (100) is performed as explained above.
The reception device shown in FIG. 42 repeats a series of operations, such as the cancelling of replicas, the equalization, the spectrum demapping, the decoding, and the generation of replicas, and thereby gradually increases the reliability of the decoded bits. After the series of operations are performed the predetermined number of times, the determining units 207 and 208 perform hard determination on bits, and then the transmission data are reproduced as decoded data.
As a system for multiplexing transmission data pieces transmitted from multiple transmission stations with use of SC-ASA, an FDMA (Frequency Division Multiple Access)-based system has been also proposed in which the point number of IDFT (Inverse Discrete Fourier Transform) is set by the transmission station to be greater than that of DFT, and subcarriers added null data are used by another transmission station (see, for example, Non-Patent Document 3).
FIGS. 44A and 44B are schematic block diagrams illustrating configurations of a transmission station device and a reception station device when user multiplexing is performed by two conventional transmission stations with use of SC-ASA. Regarding the transmission device shown in FIG. 44A, two pieces of transmission data 1 and 2 are encoded by the encoders A1000-1 and A1000-2, and the encoded transmission data pieces are modulated by modulators A1001-1 and A1001-2, respectively. The signals modulated by the modulators A1001-1 and A1001-2 are converted into parallel signals by the S/P converters A1002-1 and A1002-2, and then converted into frequency-domain signals by DFT units A1003-1 and A1003-2. Then, the frequency-domain signals are mapped by the spectral mapping units A1004-1 and A1004-2 so that the transmission data 1 and 2 are not transmitted using the same frequency subcarriers. In this case, the frequency-domain signals are mapped onto subcarriers that have good SNR (Signal to Noise Ratio) or SNIR (Signal to Noise Interference Ratio) and have frequencies not used by other users.
The mapped frequency-domain transmitted signals are converted into time-domain signals by IDFT units A1005-1 and A1005-2, and then converted into serial signals by the P/S converters A1006-1 and A1006-2. Then, cyclic prefixes are inserted into the serial signals by the CP (Cyclic Prefix) inserter A1007-1 and A1007-2. Then, the serial signals are converted into analog signals by the D/A converters A1008-1 and A1008-2. Finally, the analog signals are upconverted into radio frequency signals by radio units A1009-1 and A1009-2, and transmitted from transmission antennas 1010-1 and 1010-2.
Regarding the reception device shown in FIG. 44B, a received signal generated by multiplexing two signals simultaneously transmitted is received by a reception antenna 1100. Then, the received signal is downconverted by a radio unit A1111. The downconverted received signal is converted into a digital signal by an A/D converter A1101. Then, a cyclic prefix is removed from the digital signal by a CP (Cyclic Prefix) remover A1102. Then, the digital signal from which the cyclic prefix has been removed is converted into parallel signals by the S/P converter A1103. The parallel digital signals are converted into frequency-domain signals by a DFT unit A1104. Then, subcarriers of the respective frequency-domain signals are reversely allocated, and thereby frequency-domain signals transmitted from the respective transmission stations are demultiplexed. Then, the frequency-domain signals are independently equalized for the respective pieces of transmission data by signal equalizers A1106-1 and A1106-2, and then converted into time-domain signals by IDFT units A1107-1 and A1107-2. Then, the time-domain signals are converted into serial signals by P/S converters A1108-1 and A1108-2, and then demodulated by demodulators A1109-1 and A1109-2. Thus, decoded data 1 and 2 transmitted from the respective transmission stations are obtained from decoders A1110-1 and 1110-2.
As an equalization method performed by the signal equalizer A1106, MMSE (Minimum Mean Square Error)-based equalization is used. Generally, a tap minimizing an evaluation function J shown in an expression (107) is calculated in MMSE equalization.J=E[|WHr−s|2]  (107)
In an expression (107), E[x] denotes a mean value of x. W denotes a complex tap matrix including column vectors each being an optimal tap vector for each symbol included in DFT-S-OFDM symbols. r denotes a complex-time-domain received signal vector. s denotes a time-domain transmitted signal vector. AH denotes a Hermitian transpose of a matrix A. In this case, an optimal tap coefficient W is called a Wiener solution expressed by an expression (108).W=H(HHH+σ2I)−1  (108)
In the expression (108), H denotes a time-domain channel matrix. σ2 denotes noise variance. I denotes a unit matrix. Particularly when a frequency-domain signal operation is performed, a matrix having diagonal elements identical to frequency responses calculated by use of Fourier transform from channel impulse responses may be used as a channel matrix. Therefore, when frequency-domain received signals are used, the tap coefficients expressed by the expression (108) can be transformed as the following expression (109) where Ξ denotes channel frequency responses.W=FHΞ(ΞΞH+σ2I)−1F  (109)
In the expression (109), F denotes a matrix for performing DFT and FH denotes a matrix for performing inverse DFT. When a time-domain received signal r is multiplied by the tap matrix, the equalized received signal z can be expressed as an expression (110).z=FHΞ(ΞΞH+σ2I)−1Fr=FHΞ(ΞΞH+σ−2I)−1R  (110)
In the above expression, R=Fr, i.e., R denotes the received signal r converted by DFT into a frequency-domain signal. According to the expression (110), when a normal received signal is input and equalized in the frequency domain, the received signal is converted by DFT, multiplied by a Hermitian transpose of a matrix obtained by removing both FH and F of the expression (109), and then converted by IDFT into a time-domain signal again. Accordingly, a normal MMSE filter receives a frequency-domain received signal and a channel frequency response, and outputs a signal equalized in the frequency domain.
On the other hand, when a reception device is configured to include a canceller, such as SC/MMSE (Soft Canceller followed by MMSE), which performs repeated operation, interference waves are cancelled from received signals by use of replicas of signals generated based on the reliability of bits received from the decoder, and thereby the precision of signals input to the equalizer. Accordingly, signals input to the equalizer differ for each repletion operation. For this reason, the term corresponding to the received signal r of the evaluation function expressed by the expression (107) becomes one generated by signals other than desired signals being cancelled. Therefore, the equalized signal can be expressed as an expression (111).z=(1+γδ)−1[γsrep(k)+FHΨRrest]  (111)
In the expression (111), Rrest denotes a residual that remains without being cancelled and is generated by subtracting, from actual time-domain received signals, replicas of the received signals generated by multiplying replicas of the time-domain signals by channel characteristics. srep(k) denotes a replica of a transmitted signal for the k-th sample. γ and δ are real numbers used when tap coefficients are calculated. Similarly, Ψ denotes a complex square matrix having the size of the DFT-S-OFDM symbol length, which is used when tap coefficients are calculated. These are calculated by use of the frequency-domain channel characteristics and the replicas of frequency-domain signals (see, for example, Non-Patent Document 2). Since a replica is not input in the first operation in the expression (111) (i.e., srep(k)=0), this case is a case of the optimal tap in the expression (107), and therefore the expression (111) becomes identical to the expression (109).
Accordingly, in the case of SC/MMSE equalization, a frequency-domain residual is input as an input signal, a replica of a time-domain signal and frequency-domain channel characteristics are input, and then a time-domain signal is output. As shown in the expression (111), the operation of canceling elements other than desired elements is performed by firstly calculating the residual Rrest, and then reconfiguring the desired elements by use of the replicas of the transmitted signals and the channel characteristics. Consequently, the desired elements can be uniquely expressed among the DFT-S-OFDM symbols. Further, the same residual Rrest can be used for the DFT-S-OFDM symbols, thereby enabling a reduction in the amount of calculation including inverse matrix calculation.
[Non-Patent Document 1] “A Study on Broadband Single Carrier Transmission Technique using Dynamic Spectrum Control,” RCS 2006, January 2007.
[Non-Patent Document 2] M. Tuchler and J. Hagenauer, “Linear time and frequency domain turbo equalization,” Proc. VTC, pp. 2773-2777, Rhodes, Greece, October 2001.
[Non-Patent Document 3] “A Study on Dynamic Spectrum Control based Co-Channel Interference Suppression Technique for Multi-User MIMO Systems,” Proceedings of the IEICE General Conference in 2006, March, 2007.